Tech Tips February 2007
February 27, 2007
TECH TIP OF THE WEEK:
Use motor current to monitor load.
Every wire carrying an electric current cloaks itself in a magnetic field. You can find the direction of the field by using your right hand to grab the wire so that your thumb points in the current direction. Your fingers then wrap in the magnetic field's direction. When the ac current reverses, of course, the magnetic field reverses with it.
'A funny thing happened on the way to…,' begins a whole series of jokes, including a play, and a movie that is effectively a 99 minute running gag. For present purposes: A funny thing happened to electric power on the way from the generating station to the motor's load. What happened, or, more correctly, what happens each and every time an electric motor drives a mechanical load is that the electricity gods arrange the Universe so that the motor draws just enough current to provide the power needed to move the load at the drive's output voltage. Basically, the more resistance the load puts up, the harder the motor has to turn the crank and the more current it draws.
Why current and not voltage? Because electric motors push via magnetic fields, and currents generate magnetic fields, not voltages. The point being that motor current is the most immediate and sensitive indicator of what the motor is doing at any instant. So, if you want to see what's happening in your mechanical system, monitor the current through the motor driving it.
If anything happens to make the mechanical system work harder, such as a jam, a bad bearing, or simply trying to cram the proverbial 50 lb of stuff into the 10 lb bag, the motor current will go up. If, on the other hand, something happens to make the mechanical system easier to turn, such as a drive shaft breaking or a suction pump losing its prime, the current will go down suddenly and precipitously. Conversely, if you want to control your system to maintain a certain load power, a really good way to do it is to maintain the appropriate current level.
For example, the February Control Engineering Machine Control Monthly e-Newsletter carries a story in which a grain storage facility uses current feedback to coordinate two conveyor systems. One conveyor lifts grain out of a vessel and deposits the stuff onto another conveyor that carries it to the top of a storage silo. The second conveyor is a series of buckets that they want filled enough to use the system's capacity, but not enough to spill over on the way to the silo. The power required (neglecting bearing losses, etc.) equals the weight of the grain in each bucket times the conveyor's height times the number of buckets carrying grain simultaneously, divided by how long it takes for each bucket to go from the bottom to the top. Since the system uses an ac motor driven by a variable speed drive, the root-mean-squared (RMS) current draw equals that power divided by the drive's RMS output voltage. If the first conveyor delivers more grain, the amount in each bucket goes up, which raises the power requirement, and ultimately the current draw. If the first conveyor is a bit slow, the buckets are insufficiently filled, so the power goes down and the current draw goes down.
One type of current sensor acts as a transformer with a one-turn primary. The secondary puts out a voltage proportional to the ac current.
To maintain the correct fill level for the buckets, the system designer put a current sensor on the input to the second conveyor's motor, and fed the result back to a PLC setting the speed for the first conveyor's drive. If the bucket levels fall, the PLC sees a drop in current and reacts by raising the first conveyor's speed. If the bucket-fill-level rises, the PLC slows the first conveyor. As far as the PLC knows, it's just trying to maintain the monitored current at its setpoint. Basic physics ensures that the current level reflects the bucket fill level—as long as the conveyors are working properly. If something breaks or jams in the second conveyor, the current will drop or rise faster than is possible just by a speed mismatch between the conveyors. The PLC can be programmed to recognize such patterns, take action to bring the system safely to a stop, and set off appropriate alarms. Similar current sensors on the first conveyor's motor drive would provide instant notification of problems there.
So, how do we monitor a motor's current draw? We do it by monitoring magnetic fields surrounding the wires carrying current into the motor. Assuming, for example, that it's a three-phase motor, there will be three wires, each carrying the current for one of the phases. Ampere's Law says that each of those wires will surround itself along its full length with a magnetic field whose lines of force wrap around it the way your fingers would if you grabbed the wire with your thumb pointing along the wire, and whose strength is proportional to the current passing through the wire at that instant. In other words, for an ac drive current, the magnetic field rises and falls in step with and with a magnitude proportional to the rising, falling and reversing current through the wire.
One type of current sensor (useable only for ac) provides a ring of magnetic material (barium ferrite is probably best, although other materials work quite well, too) arranged so that the wire passes through the hole. The ring captures some of the magnetic field, which circulates around inside the ring. Ampere's Law and some interesting tensor calculus (well, some people find tensor calculus interesting) ensure that no matter how you twist and turn and wiggle the ring around, as long as the wire passes through the hole, the amount of field (magnetic flux) circulating inside depends only on the geometry and material of the ring and the instantaneous current passing through the hole. So, now we have a magnetic flux circulating around the ring, the value of which rises and falls in step with the ac current. Essentially, the sensor acts as a transformer with a one-turn primary.
Finally, the sensor manufacturer wraps a lot of turns of wire on the ring so that each turn passes through the hole and around the outside. Faraday's Law assures us that as the flux in the ring rises and falls, a voltage will appear between the coil's free ends that is proportional to the rate of change of the flux times the number of turns through the coil. Going back to calculus class, we find the peak voltage to depend on the number of turns, the ac frequency, and the peak current in the wire. While the sensor does a great job of detecting and measuring ac currents, it requires correction for the drive frequency.
That sensor does not, however, work for dc motors because, when the current level doesn't change, the flux doesn't change, and there's no voltage across the coil. While the current through a dc motor does vary somewhat as the motor turns, such changes are incidental and not to be trusted as an indicator of draw current. You need a sensor that will detect steady magnetic fields. There are a number of them, but the best such sensor for the money is the Hall effect sensor. Explaining the Hall effect is beyond the scope of this article. Let's just say that the sensor is a small block of semiconducting material carrying a current. When in the presence of a magnetic field, a voltage appears at right angles to both the current through the block and the magnetic field lines. Since the Hall voltage is proportional to the magnetic field intensity at the block, it senses the field directly, rather than its rate of change. It, therefore, doesn't care whether it's seeing ac or dc or some strange waveform, nor does it care what the ac frequency is.
Both of these sensors are relatively inexpensive, robust, and very accurate. Both provide a voltage output that can be conditioned for the input of any PLC (see the Jan. 9, 2007, Question on instrumentation amplifiers) or data acquisition front end.
For more information about current sensors, visit the Control Engineering Website , and type 'current measurement' into the search box on any page.
C.G. Masi , Control Engineering senior editor
February 20, 2007
TECH TIP OF THE WEEK:
There's no such thing as 'error,' at least according to calibration gnomes. Calibration gnomes are those folks who hide out in back rooms surrounded by all sorts of strange and often antique-looking instrumentation, and only come out into the bright factory lights to secret away your precious measurement equipment (always, seemingly, just when you need to use it) only to return it later with a new calibration sticker. In polite talk, they're called 'metrologists,' because they dedicate their lives to the science of accurate measurement.
If you say 'error' to a calibration gnome, he (or she) will respond either with a blank stare (if being polite) or a pained look that says, 'You're too ignorant to waste breath on.' If they happen to be feeling unusually sociable, they might explain patiently that 'error' is a poorly defined slang term that only uneducated people like top-flight engineers and Nobel-laureate scientists use. All-knowing metrologists use the precisely defined term 'uncertainty.'
You see, folks are always claiming that some measurement or other contains some amount of error. To a metrologist, that's rediculous. It implies that the measurement is somehow 'wrong.' A measurement (more precisely a measurement result) can't be wrong. It is whatever it is.
'Well,' the brutish engineer says, 'I mean that it differs from the 'true' value.'
'What is that?' the metrologist patiently asks, using the Socratic Method to try to inject some wisdom into the poor sot's skull.
'Unh, well, whatever the result should be ,' comes the reply.
'How would you know what that is?'
'You measure it 837,849,307,205 times and take the average,' says the engineer, brightly, vaguely remembering something from Freshman Physics lab.
'Ah,' says the metrologist, sardonicly, 'now you're getting into statistics. Statistically, there will be (depending on your instrument's precision) up to 837,849,307,205 different values showing a (hopefully) Gaussian distribution characterized by a mean square variance, which we call 'uncertainty.' So, your actual value is really only the average of a large number of uncertain measurements.'
'Yeah, but the measurement I just took typically differs from that average,' the engineer says, Swiftily.
'Yes, but if you take 47,589,493,785 readings using that meter, you'll get (again, depending on the measuring equipment's precision) up to 47,589,493,785 different values with their own Gaussian distribution, whose width would be the uncertainty in that measurement set,' the metrologist intones.
'Aha!' says the engineer, 'But that average won't match the average of the other set. That's what I mean by 'error.''
'No,' the metrologist responds, 'that is the difference between the averages of two measurement sets, one set measured with one instrument and the other with another. There's no saying which is 'right' and which is 'in error.' They are both 'uncertain.''
'But, but, but…!'
'Quiet. You sound like a motorboat,' says the metrologist. 'Our job is to make sure that the average measurements made with all the instruments in the plant do not deviate from the average measurements of certain calibration artifacts (which we keep under lock and key) by more than their specifications allow.'
'And, that's what I mean by error,' shouts the engineer!
'Oh,' says the metrologist, who then slinks off to get another cup of coffee in the cal lab—where his wisdom is more appreciated.
The best way to avoid such semantic battles with calibration gnomes is to get to know the language they use. It's also a very good way to understand what your sensors, which are the starting points for your control signals, are really telling you.
A thermometer stuck in a vat, for example, does not tell you the temperature of the goo in the vat. It tells you what its temperature is when it's sitting in the vat goo. Big difference! Helps you figure out the control system's dynamic response characteristics.
One good way to learn calibrationese is to read Measurement Uncertainty , by Ronald H. Dieck. The fourth edition, recently published by the ISA—The Instrumentation, Systems, and Automation Society, incorporates the uncertainty technologies embodied in both U.S. and international standards with a focus on understanding the strengths and weaknesses of each. The book is designed to serve as a practical desk reference in situations that commonly confront an experimenter. Key topics include the basics of the measurement uncertainty model, non-symmetrical systematic standard uncertainties, random standard uncertainties, use of correlation, curve-fitting problems, and probability plotting, combining results from different test methods, calibration errors, and uncertainty propagation for both independent and dependent error sources. Examples and problems have been included to illustrate the principles and applications of measurement uncertainty analysis.
The book is useful for test, process, and control engineers, as well as researchers, plant supervisors, managers, executives, and all others who need a basic understanding of the assessment and impact of uncertainty on measurements. For more information about Measurement Uncertainty or the rest of ISA's resources, visit www.isa.org/books .
C.G. Masi , Control Engineering senior editor
February 13, 2007
TECH TIP OF THE WEEK:
Testing surge suppressors.
Unidirectional test waveforms have a rise time of 1.2wave to decay to 50% of its peak. Source: Ferraz Shawmut.
As of Feb. 9, 2007, suppliers of UL Listed transient voltage surge suppressors (TVSSs), such as MOVs, must comply with the latest revisions of UL 1449—Standard for Transient Voltage Surge Suppressors , 2nd Edition, to maintain their UL Listing. Changes to the abnormal overvoltage tests are intended to ensure that devices are adequately protected against thermal runaway caused by the overcurrents created by overvoltage conditions.
UL 1449 identifies several test requirements for safety and performance. The tests for confirming safe operation of permanently connected TVSS products include:
Overvoltage test verifies that the TVSS will withstand an overvoltage of 110% of rated supply voltage for seven hours. The TVSS must pass this test without creating conditions that would increase the risk of fire or electric shock. Creation of holes in the enclosure or emission of flame, molten metal, or glowing or flaming particles is not allowed.
Abnormal overvoltage test verifies that the TVSS can withstand specified overvoltages without creating conditions that would increase the risk of fire or electric shock. Creation of holes in the enclosure or emission of flame, molten metal, or glowing or flaming particles is not allowed. Test voltages are based on the TVSS's voltage rating. The latest revision to UL 1449 includes additional intermediate test currents to fill holes in the previous test protocol.
Full phase voltage—short circuit current abnormal overvoltage test applies full-phase voltage across the device for up to seven hours or until the TVSS is safely disconnected from the ac supply. For example, 480 V is applied across devices rated 277 V. This test is performed with available fault currents of 100 A, 500 A, 1,000 A and the value is chosen by the manufacturer. During the test, the device will likely go into thermal runaway and will need to be safely disconnected from the circuit to pass.
Limited current abnormal overvoltage test is similar to the full phase voltage test, except that a variable resistor is adjusted to limit the short circuit test current. For permanently connected devices, four TVSSs are now tested with short circuit currents of 10 A, 5 A, 2.5 A and 0.5 A respectively. The devices are energized for up to seven hours, until the temperature of the TVSS attains equilibrium or until the TVSS is safely disconnected from the ac supply.
Measured limiting voltage test verifies that the TVSS will perform according to its marked suppressed voltage rating without any evidence of fire, operation of protective devices, or creation of openings that expose energize parts. Each TVSS is first tested with impulse surges of 6 kV and 0.5 kA, then subjected to 20 consecutive pulses of 6 kV and 3.0 kA. Ten of the pulses are positive, and ten are negative and are less than 60 seconds apart. After the TVSS is allowed to cool to room temperature, the first test with impulse surges of 6 kV and 0.5 kA are re-run. The measured limited voltage cannot deviate by more than 10% of the initial test and cannot exceed the manufacturer's marked suppressed voltage rating by 10%.
Surge current test verifies that the TVSS will withstand impulse surges of 6 kV and 10 kA without any evidence of fire, operation of protective devices, or creation of openings that expose energized parts. Each TVSS is subjected to one positive impulse and one negative impulse. After the impulse tests, the device must stay connected to an ac circuit at rated voltage.
Temperature test verifies that two-port connected TVSSs can operate at maximum rated voltage, current and frequency without adversely affecting product materials or exceeding temperature limits set forth in the standard.
Dielectric voltage-withstand test verifies that the TVSS can withstand a voltage of 1,000 V at 60 Hz, plus two times rating for one minute for various application points.
Withstand test verifies that the specified branch circuit overcurrent protection device is capable of opening short circuits downstream of the TVSS without damage to the TVSS. The test circuit is calibrated to the short circuit current marked on the TVSS. TVSSs that are able to be connected to ac power circuits and having current supplied through them to loads must be subjected to this test.
IEEE standard test waveforms
IEEE C62.45-2002, IEEE Recommended Practice of Surge Testing for Equipment Connected to Low-Voltage AC Power Circuits , identifies two waveforms that are commonly used for testing the performance of transient voltage surge suppressor (TVSS) devices:
Unidirectional . A transient waveform with a risetime 1.2rrent to rise from 10% to 90% of its peak value. Decay time is the time the wave takes to decay to 50% of its peak value. Source impedances are selected to deliver large currents (and energy) for testing TVSSs intended for such indoor locations as feeders and short branch circuits.
Oscillatory . A 100 kHz tone decaying with a decay time of 0.5rs are selected to limit current and energies to lower levels than the unidirectional waveforms.
In both these test configurations, the voltage waveform is the open-circuit voltage of the test generator and the current waveform is the current obtained by shorting the output of the test generator.
Michael J. Lang, field engineer manager, Ferraz Shawmut
This article is an edited excerpt from 'Addressing overcurrent issues for transient voltage surge suppressors.' To download the entire articles, click here .
February 6, 2007
TECH TIP OF THE WEEK:
Remember basic physics.
Control systems combined with the equipment they control behave like damped harmonic oscillators. They all overshoot when presented with a step-function control input. When underdamped, they oscillate around the setpoint, but the oscillations die out over time. At critical damping, the response error returns to zero quickly, but does not overshoot to the negative side.
In all of our talk about sophisticated control technology, it's easy to forget control systems' roots in basic physics. What any control system ultimately controls is a physical system, and physical systems obey physics.
The basic physics that control systems obey is that of a harmonic oscillator. No matter how sophisticated the system is, how many feedback loops it includes, and how long the PLC ladder code is, the built-in feedbacks and time delays follow the forced harmonic oscillator equation, which has one and only one solution:
r = A exp(-dt) sin(wt)
where r is the system's response, A is an amplitude, d is a damping factor, t is time, and w is an oscillation frequency.
If you want to see this in action, look at a video of the Wright brothers' early test flights. You'll see the aircraft bobbing up and down with a period suspiciously similar to the average human reaction time of just under a second. Aeronautical engineers who recreated the 1903 Wright flyer at the beginning of this decade learned the reason for this bobbing the first time they tried to fly it. The darn thing suffers from the dreaded pilot-induced instability (PIO) that has claimed the lives of so many test pilots.
PIO arises when the aircraft's control delays interact with the pilot's reaction time in such a way that the combined system oscillates: the pilot perceives the need for an adjustment in one direction, but by the time he or she makes the appropriate response and the aircraft reacts to that response, the motion has progressed to the point that the opposite response is needed! The pilot then makes that correction, but by the time it has an effect, it's out of phase again! With each cycle, the deviations from straight and level grows until the aircraft goes completely out of control.
The point is that controlled systems—as a whole—react as harmonic oscillators and all of the sophisticated technology involved in PID loop tuning is there to ensure that the parameters in the equation above add up to nice behavior.
Instability arises when feedbacks and delays conspire to make the damping coefficient negative. Oscillations grow rather than dying away over time.
Especially, the damping factor needs to come out to a critical value. What's a critical value? Critical damping occurs when the damping coefficient equals the oscillator's resonant frequency. The problem is that predicting that resonant frequency for a complex control system coupled with a complex physical system can be difficult, indeed.
Sometimes it's easier to find out experimentally. First of all, it helps to remember what damping is supposed to do. It's a mathematical expression of viscous force. That is, it's a force that opposes (but does not overcome) motion. Its direction is always opposite the direction of motion, and its value is proportional to the speed.
Too much damping and the system has a 'heavy' or 'sluggish' response to control inputs. Too little damping and the system oscillates immediately after a control input, but the oscillation dies out over time. Critical damping is that value which is just barely enough to stop the oscillation. That's when the system reaches its setpoint in the least amount of time.
Beware, however, when the damping factor goes negative. That is the situation with PIO. When the damping factor is negative, oscillations grow, rather than shrinking over time. While the Wright flyer suffered PIO, the negative damping factor was close enough to zero that a careful pilot could bring it under control by applying counter-intuitive control inputs. The most effective way to break up the oscillation is to apply steady control pressure, rather than trying to react to the motion.
For more information on tuning control systems, visit the Control Engineering Website at and type 'PID tuning' into the search box.
For a refresher on harmonic oscillators, visit the Hyperphysics Web page .
C.G. Masi , Control Engineering senior editor